Detecting dependence part 2 by Clifford J. Sherry, Ph.D.
In an article in the last issue (Stocks & Commodities, April 1986), I outlined a means of detecting
whether dependence existed between prices in a time series. To read this article, you should recall the
method used in that article to create price change frequency histograms and divide them into classes.
Also, to avoid repetition, the exemplary data series from that article will be used.
Once you have generated the price change frequency histogram and divided it into the appropriate
classes, like thirds, and assigned each price change to a given class, you can use this sequential list of 1's,
2's, and 3's to construct a number of potentially useful charts. I would like to describe two of these
charts—the price change density histogram, and the triggered price change histogram.
In order to construct these charts, you need to construct a series of 'bins' labeled b1, b2, b3, b4,...bn,
which represent trading days, excluding weekends and holidays. You can have as many bins as you like,
but generally 30 is enough for most purposes.